Hello DB Mandrake
I use the Delos Surround Spectacular set-up disk. It has both Stereo and Dobly test tracks that are very useful when setting up speakers. There are several tracks in particular such as the Stereo Half Left Half Right, Stereo Center Imaging that really help getting the toe in and other fine adjustments just right that make a real difference with the imaging of the system.
You also get a very nice sampler disk with some really nice music as a bonus
Rob
Attachments
I think it's more politically correct to want it to sound like the "real thing". What is the "real thing", and how well are you going to be able to reproduce it? When I saw the Moody Blues back in '82 at the Portland Coliseum, the sound was incredible. When I saw Emerson Lake and Palmer at Cobo Hall in Detroit in about '73, it was the most incredible quad sound show I've ever imagined. Eric Clapton in '87 in Oakland CA had excellent sound. Most of the rest of the concerts I've been to were anywhere from "eh" to pretty bad. Presenting sound, recording it, and playing it back are each complex challenges. I noticed that recording studios are more and more using fancy software to create imaging cues for specific sounds, both amplitude and timing related. Enya is apparently one group that uses such software relatively subtly. Other examples I found on the web gave quite an amazing imaging show. Sounds were flying all over the room with amazing clarity. It's all a matter of who does what with the sound, and how little your playback system screws that up.Is there even any good sounding live pop concert that would be worth recording? I've seen Dave Matthews in NY lately. After years of not attending any bigger live concert I thought that there would be any advances in sound quality. Unfortunately it's still the same crappy sound. Nowhere near the artistic depth a studio recording can provide. So I'm always wondering why some people want their stereo to sound like a live concert. Do they even listen with their ears or just with their eyes? At DMB they probably just smoke enough
Even though you've suggested out of phase pink noise before I hadn't got around to trying it, and I have to say I'm quite surprised at just how much cancellation there is between the speakers over such a wide frequency range, and over a range of listening positions despite the somewhat asymmetric room set-up. It's certainly quite a revealing test.
No you don't get a complete notch when facing forwards like you do when you turn your head to the side - but it doesn't matter, as that's not what we're wanting to achieve - we're not trying to achieve equal time delay from both speakers to a specific ear.
What we're trying to achieve is equal time delay from the left speaker to the left ear, and from the right speaker to the right ear. This also means that there is equal time delay from the left speaker to the right ear, and right speaker to the left ear. This means each ear receives its first arrival at the same time, and the delayed crosstalk also arrives at the same time. There is a small amount of comb filtering but it is equal in both ears, so doesn't cause an image shift.
Isn't it just an approximation of the reverberant field you would get if the speakers were in phase ? All the lobes in the room generated by a pair of speakers will reverse places with their nulls, so the total reverberant field would only be the same if the absorption/diffusion in the room was uniform.
The reverberant field with reverse phase would also drop off dramatically below the frequency where the speakers were half a wavelength apart, as lobes wouldn't be able to form. Mind you this is going to be in the bass region where all bets are off due to room modes anyway.
Interesting idea. What form of measurement are you expecting to see ?
Steady-state 1/3rd octave averaged both speakers in phase vs out of phase on the same graph, measured at the equidistant point nearest to our normal listening position ?
I think the bit I highlighted in bold says it all. When we're talking about small time errors between speakers (due to path length) it will tend to get obscured by a highly reverberant room, particularly if you're sitting at some distance as the reverberation is not phase coherent at high frequencies.
I didn't say it was impossible to get decent imaging with an error in path length to the speakers, and yes, being well away from the walls is probably helpful.
However I think you would be surprised at the further improvement possible through accurate positioning of speakers, particularly in a "dry" room where it is far more noticeable.
It's not about achieving a "perfect equilateral triangle", its about achieving symmetry along the centre line. Symmetry means if you are sitting at your normal listening location looking forward, draw an imaginary line from your head directly forward - this is the centre line.
Relative to this centre line each speaker should be exactly the same azimuth angle to the sides from the listeners perspective, but it doesn't matter exactly what that angle is. Both speakers should be toed in exactly the same amount, but it doesn't matter exactly how much. And finally each speaker should be exactly the same distance from the central listening position.
Ideally if side walls are close to the speakers there should be symmetry between speaker and side-wall distances as well, although this is not as important.
You might say that this is only relevant if you're sitting dead at the "sweet spot" but this is not true as it actually makes the sweet spot much larger if there is symmetry in the speaker set-up, and it will sound better over a wide range of listening positions, even though this seems counter intuitive.
As I said, I think its because symmetry between toe in and speaker distance is what counts - an error in distance of one speaker means that toe in is not symmetrical as viewed along the equidistant line between the speakers.
Try tweaking your speaker set-up to be exactly symmetrical relative to the listener with the distance to each speaker determined through measurement the way I described in my previous post. You might be surprised at the results.
Last edited:
The problem is that you get very different readings even when moving the mic just for a few millimeters.
Are you saying the reverberent field measurement varies every few mm, assuming that the loudspeakers were re shifted to maximize the null? (I should also mention that you must fine tune the amplifier balance knob for best null.)
I'll try some measurements in my 2 rooms this weekend. Would like to see your measurements also.
David
Yes, I suppose a pair of speakers in a room can be considered as a large two element system with considerable lobing due to their spacing. Yes, the peaks and dips swap places when you reverse the phase of one unit. In the end we are only interested in the response on the equa-distance line between them. When in phase the two units should add for all frequencies on this line. When out of phase they should cancel for all frequencies.
Having done this many times for the time allignment benefit, I've always noticed the "noise floor" that remains even when you get the distance just right for cancelation of all frequencies. The lower limit of output is the reverberent field that doesn't cancel when the direct field does.
I assume that the reverberent field spectrum is identical whether the two units are in phase or out of phase. This is likely true because no room has perfect symmetry where rays with multiple reflections can cancel to any meaningful degree.
David
Just to make sure we generate comparable data, I play pink noise and use a RTA to measure the steady state response at the listening position with a) correlated pink noise in L and R and b) correlated pink noise in L and R with reversed polarity in R.
Here's pink noise data taken with REW's RTA set to 1/48 octave, FFT length = 16384, averages = exponential 0.97 - just the satellites (HP 120Hz), no sub:
SPL(C) at listening position was around 70dB.
An externally hosted image should be here but it was not working when we last tested it.
SPL(C) at listening position was around 70dB.
Ok here are my results, and in some ways they're quite surprising.
All measurements are pink noise based 1/3rd Octave smoothed with no windowing applied, so they're overall "room response". The microphone is about 2.5 metres from each speaker and carefully adjusted to be exactly equidistant, determined by the impulse response, as well as verifying optimal cancellation at high frequencies out of phase.
The first image is just to get a sense of the overall response, where red is left, purple is right, and dark blue is both channels in phase.
Things we can see from this is that overall the two channels track, but there is a large bump just below 1Khz on the right speaker which is much closer to its wall than the left speaker, as well as the wall being un-obscured. Both speakers measure very close to the same through this region when measured at 1 metre gated, so I can only assume it's the wall reflection doing this, or some other room asymmetry..
There is also quite a difference between the two between 100-200Hz, in particular around 160Hz where one has a peak and one has a dip. I suspect this is due to comb filtering due to the unequal spacing from the side wall, or perhaps modal effects.
Dark blue is the two responses summed in phase. Across most of the range the summed response is only around 4dB greater rather than 6dB despite the mic being exactly equidistant. I can only assume the reason for this is that a large portion of the steady-state response for each individual speaker is un phase correlated room reflections which on average will add to 3dB, plus a little bit of correlated signal, giving somewhere between 3dB and 6dB.
Ok, image two is a direct comparison between Left + Right, and Left - Right. Dark blue is the same Left+Right response, with light blue Left-Right.
Some observations - below 90Hz the Left-Right response rolls off, which fits in nicely with the speakers being 1.8 metres apart which is about a half wavelength at 90Hz.
Overall the difference between in phase and out of phase is far less than I was expecting, and is greatest where the speakers are most directional - about 2-4Khz, and 6Khz upwards.
I'm quite baffled by the huge mound between 800Hz and 1.5Khz for the out of phase response, where there is hardly any difference between the in and out of phase response.
My theory for this is that the large side-wall reflection from the right speaker and wall sums together at the listening point to give cause an overall phase shift between left and right speakers of around 90 degrees at this frequency. (I'm not convinced of this though, so if anyone else has a better explanation...)
Likewise at around 360Hz where in and out of phase connection gives the same level, there is probably about 90 degrees phase shift between both speakers due to the unequal distance side-walls.
Image three is the amplitude of the out of phase response subtracted from the in phase response, essentially giving a direct to reverberant ratio referenced to the 0dB line.This gives a bit clearer picture.
Even if we ignore the anomaly around 1Khz, the ratio between in phase and out of phase response is only about 4dB on average until we get about 7Khz. This seems very low to me.
Does this mean that at my listening position I really do have a direct to reverberant ratio of only around 4-6dB ? That does fit in well with my impression that at the listening distance I'm only just barely within the "critical" distance for the room, and that the room is fairly live.
I will be interested to see other peoples measurements.
Attachments
Are you able to do a 1/3rd Octave version as well ? All the little peaks and dips can obscure the overall trends. Also are you able to do what I did where I subtracted the out of phase response from the in phase response to get a 0dB referenced direct ratio ?
Sure, here you go:
An externally hosted image should be here but it was not working when we last tested it.
Thanks.
I think that gives a much clearer idea. So it looks like you have around 10dB difference over a wide frequency range, which is about what I was hoping/expecting to measure. I think it just adds further confirmation to my belief that my room sucks
Can you give some idea of the size of the room and measurement distance from the speakers etc ?
I think that gives a much clearer idea. So it looks like you have around 10dB difference over a wide frequency range, which is about what I was hoping/expecting to measure. I think it just adds further confirmation to my belief that my room sucks
Can you give some idea of the size of the room and measurement distance from the speakers etc ?
Last edited:
The room is 10.63m x 4.47m. 1 pixel equals 1cm:
Red = satellites, blue = subs. The whole front wall is damped with 20cm polyester batting + 20cm air gap.
An externally hosted image should be here but it was not working when we last tested it.
Red = satellites, blue = subs. The whole front wall is damped with 20cm polyester batting + 20cm air gap.
So I take it the top curve is both units in phase, second curve is units cancelling and 3rd is noise floor?
Reverberent field is about 10dB below direct field. That puts you inside the critical distance.
Thanks for showing.
David
Of course the method relies on cancelation to reveal the reverberent field. Cancelation requires balanced levels. Your system seems well balanced but there are some room effects giving curve differences in the 70 to 300 range. For others doing the method, try and null out the difference curve with some fine tuning of the system balance knob.
Of course if room effects are giving different curves then that should be considered part of the reverberent field (early reflections).
It looks to me that you are just about at the critical distance: reverberent level is just a little below the direct level. My understanding is that most domestic system have listeners about at the critical distance so you are in good company.
Thanks for the curves, wish I didn't have to work today.
Regards,
David
Yes.
That was my goal. Now if someone would build an asymmetrical waveguide so I could have some time/level trading...
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.